[llvm] [mlir] [NVPTX] Added more MMA intrinsics for F8F6F4 and FP64 types. (PR #156040)

[Artem Belevich via llvm-commits]

================
@@ -90,6 +90,21 @@ def __init__(self, geom, frag, ptx_elt_type, is_mma_sparse=False):
             "m16n8k32:b:s8": 2,
----------------
Artem-B wrote:

Good news and bad news.

Good news is that there's already a wmma-ptx87-sm120a.py test variant that will generate test cases for sm_87 and sm_120a. I've missed it when I wrote the previous comment.

https://github.com/llvm/llvm-project/blob/7c5b535d8c50e52ba13439ce52b9fe6fa34afcab/llvm/test/CodeGen/NVPTX/wmma-ptx87-sm120a.py#L5

The bad news is that the test has not been updated to use new way lit specifies ptxas features, so the ptxas tests we not actually run. You do need to update that test with `%if ptxas-sm_120a && ptxas-isa-8.7` and rerun that test. No need to include the generated files.
 

https://github.com/llvm/llvm-project/pull/156040